Machine Learning-Assisted Eu(III)-Functionalized HOF-on-HOF Composite-Based Sensor Platform for Precise and Visual Identification of Multiple Pesticides.

Abstract

Precise and rapid identification of pesticides is crucial to ensure a green environment, food safety, and human health. However, complex sample environments often hinder precise identification, especially for simultaneous differentiation of multiple pesticides. Herein, we first synthesize a Eu(III)-functionalized HOF-on-HOF composite (Eu@PFC-1@MA-TPA) and then utilize principal component analysis (PCA) and a machine learning (ML) algorithm to achieve simultaneous identification of the pesticides 2,6-dichloro-4-nitroaniline (DCN) and thiabendazole (TBZ) and their mixtures. Eu@PFC-1@MA-TPA displays high quantitative identification ability, which can distinguish single DCN and TBZ as low as 1 μM and their mixtures at 5 μM through PCA. In addition, the hydrogel film Eu@PFC-1@MA-TPA/AG is fabricated to monitor DCN and TBZ in drinking water, tap water, river water, and apple juice with high sensitivity. Furthermore, based on the obvious fluorescence color variance of pesticides, Eu@PFC-1@MA-TPA/AG achieves visual and in situ imaging detection of single DCN and TBZ and their mixtures. More importantly, we construct an intelligent artificial vision platform integrating Eu@PFC-1@MA-TPA/AG with a DenseNet algorithm, which can identify the concentrations and types of DCN and TBZ and their mixtures within 1 s with over 98% accuracy. This work develops a precise and rapid analysis method for simultaneous identification of multiple pesticides through combining a visualized fluorescence sensor and an ML algorithm.